Prediction on overlay smoothness improvements using Artificial Neural Network (ANN) approach

This paper describes the ANN approach applied to the prediction of after-overlay pavement smoothness for asphalt concrete pavement. The ANN input variables used were (i) original (preoverlay) pavement roughness; (ii) overlay thickness (application rate); (iii) type of overlay material; (iv) pavement functional classification; and (v) whether or not milling was performed. The difference between pre-overlay and after-overlay smoothness is predicted by ANN output. A dataset was obtained from historical records maintained by the South Carolina Department of Transportation. The dataset was applied for neural network training and testing. After the network learned from the training dataset, new data from the similar domain were predicted. The influences of different ANN configurations and the size of the training dataset on ANN predictive quality were analyzed by various techniques. Acceptable predictive qualities were demonstrated. The study shows that ANN can be a useful mathematical tool for prediction of overlay smoothness.


  • English

Media Info

  • Pagination: 261-9
  • Monograph Title: Proceedings of the International Conference on Highway Pavement Data, Analysis and Mechanistic Design Applications, September 7-10 2003, Columbus, Ohio: volume 2

Subject/Index Terms

Filing Info

  • Accession Number: 01390915
  • Record Type: Publication
  • Source Agency: ARRB
  • Files: ATRI
  • Created Date: Aug 23 2012 5:08AM